- Email: [email protected]
Cancer patients and characteristics of pulmonary embolism
European Journal of Radiology 69 (2009) 478–482
Cancer patients and characteristics of pulmonary embolism U. Hasenberg a , T. Paul a , A. Feuersenger b , M. Goyen c , K. Kr¨oger d,∗ a
b
Department of Radiology, University Hospital Essen, Germany Institute of Medical Informatics, Biometry and Epidemiology, University Hospital Essen, Germany c Department of Radiology, University Medical Center Hamburg-Eppendorf, Germany d Department of Angiology, University Hospital Essen, Germany
Received 18 April 2007; received in revised form 12 November 2007; accepted 13 November 2007
Abstract Objective: To check the hypothesis that cancer patients suffer from extended pulmonary embolism (PE) more frequently than patients without cancer we analysed PEs proved by computed tomography (CT)-imaging. Patients and methods: One hundred and fifty consecutive CT scans at the University Hospital of Essen from March 2002 until December 2004 which proved a definite case of pulmonary embolism were retrospectively reviewed (79 men, 71 women; mean age 57 ± 15 years). Underlying disease and blood parameters were included (haemoglobin, haematocrit, fibrinogen and total protein, if determined within 48 h before the CT scans). Results: Patients with malignant disease were older (59 ± 12 years vs. 54 ± 19 years, p = 0.05) and tend to have a higher rate of central PEs (52% vs. 34%, p = 0.08) than patients without malignancies. The odds of a central PE in cancer patients was about twice as high as in patients without a malignant disease (Odds ratio: 2.08, 95%-confidence interval: 1.06–4.10; age-adjusted Odds ratio 1.88, 95%-confidence interval: 0.92–3.84). Additional adjustment for the clinical information dyspnoea, inhospital patient and clinically expected PE did not deteriorate the odds. Thrombus density determined in patients with central PE only shows a trend towards a lower density in patients with malignant disease (52 ± 13 HE vs. 45 ± 15 HE, p = 0.13). There is no statistical evidence that thrombus density is related to one of the blood parameters or even blood density measured in the pulmonary artery. Conclusion: Although this is a retrospective study including a small number of patients it shows that cancer patients are at a higher risk for central PE than patients without cancer. Characteristics of the intrapulmonal thrombus in cancer and non-cancer patients seem to be different. © 2007 Elsevier Ireland Ltd. All rights reserved. Keywords: Pulmonary embolism; CT scan; Thrombus density; Cancer
1. Introduction Since Trousseau the association of thromboembolic complications and cancer is well known. Changes in the coagulability and rheology of the blood are supposed to cause the increased frequency of deep vein thrombosis (DVT) in cancer. Procoagulant activity generated by tumor cells, macrophages, platelets and vascular endothelial cells contributes to clot formation in cancer patients. The risk of thromboembolism is influenced by the kind of cancer, the stage of the disease and the co-morbidity of the patients [1]. ∗ Corresponding author at: Klinik und Poliklinik f¨ ur Angiologie, Universit¨at Essen, Hufelandstraße 55, 45122 Essen, Germany. Tel.: +49 201 723 2530; fax: +49 201 723 5967. E-mail address: [email protected] (K. Kr¨oger).
0720-048X/$ – see front matter © 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2007.11.022
Probably due to this procoagulopathic state not only the frequency of thrombosis in cancer patients is higher than in non-cancer patients, but also DVTs are more extended. In a retrospective analysis including 489 consecutive patients with thrombosis we had shown proximal DVTs to be more frequent in cancer patients than in patients without [2]. This finding is supported by others, who analysed the thrombus size and reported that the largest thrombi were found in patients with malignancy [3]. Extended proximal DVT are associated with higher rates of pulmonary embolism in older studies (PE). PE rates of 77% and 67% were reported, when pelvic veins or thigh veins were involved, and of 46% if lower leg veins were involved [4]. Thus, we hypothezised that, if cancer patients have a higher rate of proximal DVTs compared to patients without cancer, they should also have more extended PE. To check the hypothesis that cancer patients more frequently suffer from extended PE than
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
479
patients without cancer in a retrospective design, we analysed PEs proved by computed tomography (CT)-imaging. 2. Material and methods For this study CT scans saved in the PACS system of the Department of Radiology performed from March 2002 to December 2004 at the University Hospital of Essen have been reviewed, and all scans which proved a definite case of PE were evaluated. The scans of 150 consecutive patients proving new detected PEs for the first time were finally included in our analysis. The criterion used to diagnose pulmonary emboli consisted of direct visualisation of a thrombus (central filling defect completely or partially outlined by contrast agent) or of complete occlusion by thrombus in normal-sized or enlarged vessels. In the analysis we could enrol 150 patients with PE (79 men and 71 women; age ranged from 15 to 83, mean age was 57 years and standard deviation 15.4 years). Ninety-one patients suffered from malignant disease. Patients were divided into those with central and peripheral PE by two radiologists blinded for the clinical data as underlying disease and laboratory values. Criterion for being assigned to the “central-group” was a filling defect in the pulmonary arteries or the main stem. Patients who showed filling defects in central and peripheral vessels were also assigned to the “central-group”. All other patients with filling defects in one or more peripheral vessels were assigned to the “peripheral group”. Furthermore, the blood parameters haemoglobin, haematocrit, fibrinogen and total protein content were retrospectively derived from the central laboratory. Only parameters determined within 48 h before the CT scans were performed were used for the presented analysis. Thus these parameters were available only for 101 up to 125 patients. Sonographies performed to look for DVT either 24 h before or after the CT-examination were available in 80 patients. The extension of the DVT was taken from the sonography protocol as classified as distal (up to popliteal vein), femoral (up to the inflow of the long superficial vein) and iliacal (proximal to the inflow). Due to the retrospective design of the study clinical scores as the Wells score could not be determined. Data regarding inhospital treatment and dyspnoea at the time when symptoms occurred and the differentiation between expected and unexpected accidental PEs were derived for the medical records or the ordering form. 3. Data acquisition As not all patients were scanned under the suspicion of PE, different types of scans were performed: CT angiography with 1 mm sections (n = 120), CT thorax scans with 5 mm sections with contrast material (n = 29) and heart CT scan (n = 1). Two different CT units were used: Siemens Sensation 16 and Siemens Volume Zoom (Siemens Medical Solutions, Forchheim, Germany). In all cases images were acquired in a caudocranial direction, starting at the level of the lower hemidiaphragm and ending at the top of the aortic arch.
Fig. 1. Area of measurement of dye enhanced blood density in the pulmonary artery [3] and of the central embolus itself [1,2].
The CT scans for all patients were assessed using parameters derived on the axial plane. We measured the maximum minor axis of the right ventricle and the maximum minor axis of the left ventricle this was usually at the plane of the mitral valve. The right ventricle (RV) was considered dilated if the RV cavity was wider than the left ventricle (LV) cavity along the short axis. We also measured the transversal diameter of the central pulmonary artery (on the CT image passing through the pulmonary trunk just before its branching off into left and right pulmonary arteries). For extensive central PE the thrombus density was measured as well (Fig. 1). To avoid errors by including contrast material in the measurement of the thrombus’ Hounsfield units (HU) we kept a distance of at least 2 mm to its margin. The square left was defined as the region of interest and had to have a size of at least 5 mm2 to be included. Multiple lesions in one slice were all measured, if fitting to criteria just mentioned. We did this in a 3 mm interval if possible (for 29 images with 5 mm sections we measured every section), and the mean of all these values was defined as the mean thrombus density. 4. Statistics One-way-ANOVA were used to calculate F-tests which were applied for comparison of age, PA-diameter, RV/LV ratio, Fibrinogen, Hematocrit, Hemoglobin and total protein between the groups of central and peripheral PE and between the groups of cancer and non-cancer patients. Logistic regression was performed to calculate Wald Chi-Square statistics to test for differences between the prevalences of males, thrombosis of the distal, femoral and iliacal veins, cancer and central PE, respectively. All analyses were repeated adjusting for age. p-Values of 0.05 and less were considered significant. No adjustment for multiple testing was applied because the statis-
480
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
Table 1 Characteristics of patients separated by peripheral and central pulmonary embolism Peripheral PE
Central PE
Fraction Males Inpatients Dyspnoea Suspected PE Distal-DVTb Femoral-DVTb Iliacal-DVTb
%
45/83 45/83 64/83 69/83 12/45 7/45 11/45
Fraction
54 54 77 84 27 16 24
34/67 34/67 55/67 48/67 13/35 9/35 9/35
Peripheral PE
a b c
53 29 1.1 467 35 12 6
± ± ± ± ± ± ±
%
Un-adjusted
Age-adjusted
51 51 82 73 37 26 26
0.67a
0.54a 0.41a 0.66a 0.27a 0.16a 0.37a 0.68a
0.67a 0.45a 0.09a 0.32a 0.26a 0.90a
Central PE
Mean ± S.D. Age (years) PA-diameter (mm) RV/LV ratio Fibrinogen (mg/dl)b Hematocrit (%)b Hemoglobin (g/dl)b Total protein (g/dl)b
p-Values
16 4 0.3 245 6 2 2
n
Mean ± S.D.
83 83 83 61 65 65 50
62 30 1.2 495 34 11 6
± ± ± ± ± ± ±
p-Values n
13 4 0.4 204 5 2 1
0.0002c 0.12c 0.25c 0.53c 0.52c 0.53c 0.57c
67 67 59 52 60 60 51
– 0.52c 0.53c 0.64c 0.33c 0.35c 0.42c
p-Values derived from logistic regression. Denominators of fractions respective numbers of observations (n) per row do not add to 150 due to missing data (see Section 2). p-Values derived from ANOVA.
tical analysis was performed in an exploratory manner. Odds ratios were estimated from four adjusted multivariate logistic regression models (adjusted for age, age and haemoglobin, age and hematocrit, age and fibrinogen) for central PE and malignancy. Finally an additional adjustment was done including the clinical information (inhospital, dyspnoea, expected PE).
The influence of fibrinogen, hematocrit, hemoglobin and total protein on thrombus density was analysed in a linear regression model. Pearson correlation coefficients were used to exclude a linear dependency. The same procedures were used to prove the effects of these variables on blood density. Pearson correlation coefficients were estimated to exclude correlation of blood and thrombus density. All calculations were
Table 2 Characteristics of patients with cancer and without Variables
Cancer
Non-cancer
Fraction Males Inpatients Dyspnoea Suspected PE Distal-DVTb Femoral-DVTb Iliacal-DVTb Central PEb
%
52/91 53/91 84/91 65/91 12/43 11/43 10/43 47/91
Fraction
57 58 92 72 28 26 23 52
Age (years) PA-diameter (mm) RV/LV ratio Fibrinogen, (mg/dl)b Hematocrit (%)b Hemoglobin (g/dl)b Total protein (g/dl)b Thrombus density (HU)b a b c
59 29 1.1 503 34 11 6 45
± ± ± ± ± ± ± ±
12 4 0.3 251 5 2 2 15
%
27/59 26/59 35/59 52/59 13/37 5/37 10/37 20/59
Cancer Mean ± S.D.
p-Values Un-adjusted
46 44 59 90 35 14 27 34
Mean ± S.D.
91 91 91 68 78 78 67 41
54 29 1.2 446 37 12 6 52
± ± ± ± ± ± ± ±
19 5 0.4 180 6 2 1 13
0.20a 0.13a <0.0001a 0.02a 0.64a 0.24a 0.84a 0.08b
0.09a <0.0001a 0.01a 0.49a 0.18a 0.70a 0.03b
Non-cancer n
Age-adjusted
0.17a
p-Values n 59 59 59 45 47 47 34 16
p-Values derived from logistic regression. Denominators of fractions respective numbers of observations per row do not add to 150 due to missing data (see Section 2). p-Values derived from ANOVA.
0.05c 0.37c 0.94c 0.19c 0.005c 0.006c 0.43c 0.13c
– 0.68c 0.82c 0.24c 0.002c 0.003c 0.50c 0.13c
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
done using the SAS version 9 (SAS Institute, Inc., Cary, NC, USA). 5. Results Table 1 shows the characteristics of patients with central (n = 83) and peripheral PEs (n = 67). Patients with central PE were older than patients with peripheral PEs, but there was no trend to more proximal vein thrombosis in patients with central PEs. The analysed blood parameters did not show any differences. Separating for malignant disease showed that patients with malignant disease were older and had a higher rate of central PEs than patients without malignancies (Table 2). Although dyspnoea was present in most of the cancer patients a lower rate of cancer patients were suspected to have PE compared to patients without cancer. The odds of a central PE in cancer patients was about twice as high as in patients without a malignant disease (Odds ratio: 2.08, 95%-confidence interval: 1.06–4.10; age-adjusted Odds ratio 1.88, 95%-confidence interval: 0.92–3.84; age and haemoglobin adjusted: 2.13 (0.96–4.75); age and hematocrit adjusted: 2.13 (0.95–4.75); age and fibrinogen adjusted: 2.09 (0.93–4.68). After additional adjustment for the clinical information the odds did not deteriorate (Odds ratio: 2.11, 95%-confidence interval: 0.99–4.54; age-adjusted Odds ratio 1.99, 95%-confidence interval: 0.90–4.39; age and haemoglobin adjusted: 2.23 (0.93–5.39); age and hematocrit adjusted: 2.23 (0.93–5.38); age and fibrinogen adjusted: 2.30 (0.95–5.56). In addition cancer patients had significant lower haemoglobin and hematocrit levels. Thrombus density could only be determined in patients with central PE and despite this reduction in total number there was a trend towards a lower thrombus density in patients with malignant disease. Therefore, the influence of the blood parameter on thrombus density was analysed separately. Hematocrit and haemoglobin showed a strong correlation (Pearson correlations coefficient: 0.97447), therefore only haemoglobin was included in a linear regressions model. There is no statistical evidence that the blood parameters haemoglobin, total protein or fibrinogen are related to thrombus density. Thrombus density did not correlate with blood density measured in the perfused pulmonal artery (Pearson correlation coefficient: −0.09590). 6. Discussion The results support the hypothesis that cancer patients are at a higher risk to develop central PEs. Density of the intrapulmonary thrombus seems to differ between patients with cancer and without. The increased risk for PE and PE associated mortality in cancer patients is well known [5,6]. Even the percentage of patients with venous thromboembolic disease at the initial hospitalization is higher for those with cancer compared with those without [7]. Thus in a retrospective examination in a university hospital with high cancer rates the number of patients with cancers with an embolus must be higher. In cancer patients unexpected acci-
481
dental PEs are more frequently found in routine CT scans for staging than in any other group of patients [8–10]. In accordance with these data cancer patients were more frequent in our population, too. The risk of cancer increases with age. Thus the higher age in PE-patients suffering from cancer is not an extraordinary finding and statistical analyses were adjusted for age. There are also some data that reported cancer patients to have a higher rate of massive PE [11]. Fatal PE is one of the major mortality reasons in cancer patients in post-mortem studies [1]. Specific analyses of the rates of central compared to peripheral PEs in cancer patients are not available to our knowledge. Central PE and peripheral PE do not necessarily differ in clot load, because a large embolus floated into the pulmonary artery can be fragmented and cause multiple segmental or subsegmental occlusions. In an older study proximal DVTs have been described to be associated with a higher risk of PE than distal DVT [4]. A more recent study using CT scan on 159 consecutive PE symptom-free patients with symptomatic DVT did not describe a significant association between the level of DVT and the presence of PE [12]. PE recurrences determined in 23 out of 348 patients by lung scan did not depend on proximity of DVT [13]. Even isolated soleal vein thrombosis can be found in patients with PE [14]. All these studies did not distinguish central and peripheral PE. Although we do not have information about the extension of the peripheral DVT our data do not suggest that patients with central PE suffer from more proximal DVT compared to patients with peripheral PE. From a hemodynamical point of view the clinical course of central PE often is more dramatically than in patients with peripheral PE. Although echocardiography is currently used to detect RV dysfunction, CT pulmonary angiography could replace this technique because it can be used to diagnose PE and assess RV function simultaneously. The most sensitive parameter seems to be the ratio of the right and left ventricle (RV/LV ratio). Ghaye et al. reported RV/LV ratios from 1.3 ± 0.4 in PE-survivors to 1.8 ± 0.6 in non-survivors (p = 0.011). Diameter of the pulmonary artery did not distinguish these groups [15]. The RV/LV ratios in this study with patients treated in an intensive care unit were higher than in our population in average and in other studies [16,17]. In these patients with an obviously less hemodynamical deterioration RV/LV ratio is not such sensitive. None of these studies looked for differences in central PE versus peripheral PE which is just an anatomical description. Whereas peripheral PE can have a high clot load, clinically they can be as severe as central PE. Because we did not correlate severity of PE based on clinical findings, we found no difference in RV/LV ration associated to central or peripheral PE. With the direct visualisation of the intrapulmonal embolus multidetector computed tomography allows a precise characterisation of the embolus. HU generally depend on the amount of fluid and cellular structure in the tissue. There is no valid information about HU characteristic for peripheral venous thrombus or even intrapulmonary embolus. The sequential change in density of subdural hematomas (SDHs) in CT has been retrospectively investigated in 446 cases [18]. The study showed that
482
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
density of acute SDH is usually hyperdense and becomes hypodense within 3 weeks. Then the density progressively increases by the repeated microhemorrhage, which is the mechanism of enlargement of chronic SDH. The density of chronic SDH increases with time up to 90 days, then decreases again after maturation of the neo-membrane, which is the mechanism of spontaneous resolution. In our study the differences in thrombus density could not be explained by cancer specific blood alterations as lower haemoglobin and hematocrit or higher fibrinogen. Thus unknown thrombus age or unknown mechanisms influencing the development of each single thrombus might predict thrombus density. 7. Clinical relevance CT scans can not only detect PE, but can also classify the embolus. The finding of a central PE and a lower thrombus density might represent a specific composition of the thrombus. Such composition might be due to the increased procoagulant activity associated with cancer. Differences in thrombus composition might influence thrombus organisation and solution, respectively, and might influence patient prognosis. Especially the mechanisms causing the differences in size and density of the embolus should be analysed in future studies. 8. Limitations Several limitations in our study warrant consideration. First, its retrospective design prevented us from producing correlations with clinical findings of other diagnostic tests, especially echocardiography. Second, we did not focus on a group of specific symptoms or specific severity. Thus, the rate of central PE can be different in other populations. Third, we did not know the exact age of the DVT. So we cannot correlate thrombus density to thrombus age. Forth, patient selection might have influenced the results. Cancer patients could have had more pronounced clinical symptoms of PE than patients without cancer. Clinical scores such as the Wells score have not been evaluated but would have given crucial information regarding this potential source of error. Although dyspnoea was more frequently in cancer patients the number of patients a PE was suspected in was lower in cancer patients. 9. Conclusion Although this is a retrospective study including a small number of patients it shows different characteristics of PE in cancer and non-cancer patients.
References [1] Lee AY, Levine MN. Venous thromboembolism and cancer: risks and outcomes. Circulation 2003;107:I17–21. [2] Kr¨oger K, Fischer A, Hinrichs A, Gocke C, Rudofsky G. Patients with a malignant tumor have more extended thrombosis than patients without. Angiology 1998;49:923–8. [3] Vucic N, Magdic T, Krnic A, Vcev A, Bozic D. Thrombus size is associated with etiology of deep venous thrombosis—a cross-sectional study. Coll Antropol 2005;29:643–7. [4] Mostbeck A, Partsch H, Kohn H, Konig B. Pulmonary embolism in leg and pelvic vein thrombosis. Results of a prospective study—diagnosis, frequency, nuclearmedical and clinical appearance. Wien Klin Wochenschr 1980;92:464–71. [5] Svendsen E, Karwinski B. Prevalence of pulmonary embolism at necropsy in patients with cancer. J Clin Pathol 1989;42:805–9. [6] Engelke C, Rummeny EJ, Marten K. Pulmonary embolism at multi-detector row CT of chest: one-year survival of treated and untreated patients. Radiology 2006;239:563–75. [7] Levitan N, Dowlati A, Remick SC, et al. Rates of initial and recurrent thromboembolic disease among patients with malignancy versus those without malignancy. Risk analysis using Medicare claims data. Medicine (Baltimore) 1999;78:285–91. [8] Engelke C, Manstein P, Rummeny EJ, Marten K. Suspected and incidental pulmonary embolism on multidetector-row CT: analysis of technical and morphological factors influencing the diagnosis in a cross-sectional cancer centre patient cohort. Clin Radiol 2006;61:71–80. [9] Perrier A, Roy PM, Sanchez O, et al. Multidetector-row computed tomography in suspected pulmonary embolism. N Engl J Med 2005;352:1760–8. [10] Sebastian AJ, Paddon AJ. Clinically unsuspected pulmonary embolism—an important secondary finding in oncology CT. Clin Radiol 2006;61:81–5. [11] Shen VS, Pollak EW. Fatal pulmonary embolism in cancer patients: Is heparin prophylaxis justified? South Med J 1980;73:841–3. [12] Lopez-Beret P, Pinto JM, Romero A, Orgaz A, Fontcuberta J, Oblas M. Systematic study of occult pulmonary thromboembolism in patients with deep venous thrombosis. J Vasc Surg 2001;33:515–21. [13] Monreal M, Lafoz E, Ruiz J, Callejas JM, Arias A. Recurrent pulmonary embolism in patients treated because of acute venous thromboembolism: a prospective study. Eur J Vasc Surg 1994;8(September (5)):584–9. [14] Ohgi S, Tachibana M, Ikebuchi M, Kanaoka Y, Maeda T, Mori T. Pulmonary embolism in patients with isolated soleal vein thrombosis. Angiology 1998;49:759–64. [15] Ghaye B, Ghuysen A, Willems V, et al. Severe pulmonary embolism: pulmonary artery clot load scores and cardiovascular parameters as predictors of mortality. Radiology 2006;239:884–91. [16] Araoz PA, Gotway MB, Trowbridge RL, et al. Helical CT pulmonary angiography predictors of in-hospital morbidity and mortality in patients with acute pulmonary embolism. J Thorac Imaging 2003;18:207–16. [17] van der Meer RW, Pattynama PM, van Strijen MJ, et al. Right ventricular dysfunction and pulmonary obstruction index at helical CT: prediction of clinical outcome during 3-month follow-up in patients with acute pulmonary embolism. Radiology 2005;235:798–803. [18] Lee KS, Bae WK, Bae HG, Doh JW, Yun IG. The computed tomographic attenuation and the age of subdural hematomas. J Korean Med Sci 1997;12:353–9.
Cancer patients and characteristics of pulmonary embolism U. Hasenberg a , T. Paul a , A. Feuersenger b , M. Goyen c , K. Kr¨oger d,∗ a
b
Department of Radiology, University Hospital Essen, Germany Institute of Medical Informatics, Biometry and Epidemiology, University Hospital Essen, Germany c Department of Radiology, University Medical Center Hamburg-Eppendorf, Germany d Department of Angiology, University Hospital Essen, Germany
Received 18 April 2007; received in revised form 12 November 2007; accepted 13 November 2007
Abstract Objective: To check the hypothesis that cancer patients suffer from extended pulmonary embolism (PE) more frequently than patients without cancer we analysed PEs proved by computed tomography (CT)-imaging. Patients and methods: One hundred and fifty consecutive CT scans at the University Hospital of Essen from March 2002 until December 2004 which proved a definite case of pulmonary embolism were retrospectively reviewed (79 men, 71 women; mean age 57 ± 15 years). Underlying disease and blood parameters were included (haemoglobin, haematocrit, fibrinogen and total protein, if determined within 48 h before the CT scans). Results: Patients with malignant disease were older (59 ± 12 years vs. 54 ± 19 years, p = 0.05) and tend to have a higher rate of central PEs (52% vs. 34%, p = 0.08) than patients without malignancies. The odds of a central PE in cancer patients was about twice as high as in patients without a malignant disease (Odds ratio: 2.08, 95%-confidence interval: 1.06–4.10; age-adjusted Odds ratio 1.88, 95%-confidence interval: 0.92–3.84). Additional adjustment for the clinical information dyspnoea, inhospital patient and clinically expected PE did not deteriorate the odds. Thrombus density determined in patients with central PE only shows a trend towards a lower density in patients with malignant disease (52 ± 13 HE vs. 45 ± 15 HE, p = 0.13). There is no statistical evidence that thrombus density is related to one of the blood parameters or even blood density measured in the pulmonary artery. Conclusion: Although this is a retrospective study including a small number of patients it shows that cancer patients are at a higher risk for central PE than patients without cancer. Characteristics of the intrapulmonal thrombus in cancer and non-cancer patients seem to be different. © 2007 Elsevier Ireland Ltd. All rights reserved. Keywords: Pulmonary embolism; CT scan; Thrombus density; Cancer
1. Introduction Since Trousseau the association of thromboembolic complications and cancer is well known. Changes in the coagulability and rheology of the blood are supposed to cause the increased frequency of deep vein thrombosis (DVT) in cancer. Procoagulant activity generated by tumor cells, macrophages, platelets and vascular endothelial cells contributes to clot formation in cancer patients. The risk of thromboembolism is influenced by the kind of cancer, the stage of the disease and the co-morbidity of the patients [1]. ∗ Corresponding author at: Klinik und Poliklinik f¨ ur Angiologie, Universit¨at Essen, Hufelandstraße 55, 45122 Essen, Germany. Tel.: +49 201 723 2530; fax: +49 201 723 5967. E-mail address: [email protected] (K. Kr¨oger).
0720-048X/$ – see front matter © 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2007.11.022
Probably due to this procoagulopathic state not only the frequency of thrombosis in cancer patients is higher than in non-cancer patients, but also DVTs are more extended. In a retrospective analysis including 489 consecutive patients with thrombosis we had shown proximal DVTs to be more frequent in cancer patients than in patients without [2]. This finding is supported by others, who analysed the thrombus size and reported that the largest thrombi were found in patients with malignancy [3]. Extended proximal DVT are associated with higher rates of pulmonary embolism in older studies (PE). PE rates of 77% and 67% were reported, when pelvic veins or thigh veins were involved, and of 46% if lower leg veins were involved [4]. Thus, we hypothezised that, if cancer patients have a higher rate of proximal DVTs compared to patients without cancer, they should also have more extended PE. To check the hypothesis that cancer patients more frequently suffer from extended PE than
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
479
patients without cancer in a retrospective design, we analysed PEs proved by computed tomography (CT)-imaging. 2. Material and methods For this study CT scans saved in the PACS system of the Department of Radiology performed from March 2002 to December 2004 at the University Hospital of Essen have been reviewed, and all scans which proved a definite case of PE were evaluated. The scans of 150 consecutive patients proving new detected PEs for the first time were finally included in our analysis. The criterion used to diagnose pulmonary emboli consisted of direct visualisation of a thrombus (central filling defect completely or partially outlined by contrast agent) or of complete occlusion by thrombus in normal-sized or enlarged vessels. In the analysis we could enrol 150 patients with PE (79 men and 71 women; age ranged from 15 to 83, mean age was 57 years and standard deviation 15.4 years). Ninety-one patients suffered from malignant disease. Patients were divided into those with central and peripheral PE by two radiologists blinded for the clinical data as underlying disease and laboratory values. Criterion for being assigned to the “central-group” was a filling defect in the pulmonary arteries or the main stem. Patients who showed filling defects in central and peripheral vessels were also assigned to the “central-group”. All other patients with filling defects in one or more peripheral vessels were assigned to the “peripheral group”. Furthermore, the blood parameters haemoglobin, haematocrit, fibrinogen and total protein content were retrospectively derived from the central laboratory. Only parameters determined within 48 h before the CT scans were performed were used for the presented analysis. Thus these parameters were available only for 101 up to 125 patients. Sonographies performed to look for DVT either 24 h before or after the CT-examination were available in 80 patients. The extension of the DVT was taken from the sonography protocol as classified as distal (up to popliteal vein), femoral (up to the inflow of the long superficial vein) and iliacal (proximal to the inflow). Due to the retrospective design of the study clinical scores as the Wells score could not be determined. Data regarding inhospital treatment and dyspnoea at the time when symptoms occurred and the differentiation between expected and unexpected accidental PEs were derived for the medical records or the ordering form. 3. Data acquisition As not all patients were scanned under the suspicion of PE, different types of scans were performed: CT angiography with 1 mm sections (n = 120), CT thorax scans with 5 mm sections with contrast material (n = 29) and heart CT scan (n = 1). Two different CT units were used: Siemens Sensation 16 and Siemens Volume Zoom (Siemens Medical Solutions, Forchheim, Germany). In all cases images were acquired in a caudocranial direction, starting at the level of the lower hemidiaphragm and ending at the top of the aortic arch.
Fig. 1. Area of measurement of dye enhanced blood density in the pulmonary artery [3] and of the central embolus itself [1,2].
The CT scans for all patients were assessed using parameters derived on the axial plane. We measured the maximum minor axis of the right ventricle and the maximum minor axis of the left ventricle this was usually at the plane of the mitral valve. The right ventricle (RV) was considered dilated if the RV cavity was wider than the left ventricle (LV) cavity along the short axis. We also measured the transversal diameter of the central pulmonary artery (on the CT image passing through the pulmonary trunk just before its branching off into left and right pulmonary arteries). For extensive central PE the thrombus density was measured as well (Fig. 1). To avoid errors by including contrast material in the measurement of the thrombus’ Hounsfield units (HU) we kept a distance of at least 2 mm to its margin. The square left was defined as the region of interest and had to have a size of at least 5 mm2 to be included. Multiple lesions in one slice were all measured, if fitting to criteria just mentioned. We did this in a 3 mm interval if possible (for 29 images with 5 mm sections we measured every section), and the mean of all these values was defined as the mean thrombus density. 4. Statistics One-way-ANOVA were used to calculate F-tests which were applied for comparison of age, PA-diameter, RV/LV ratio, Fibrinogen, Hematocrit, Hemoglobin and total protein between the groups of central and peripheral PE and between the groups of cancer and non-cancer patients. Logistic regression was performed to calculate Wald Chi-Square statistics to test for differences between the prevalences of males, thrombosis of the distal, femoral and iliacal veins, cancer and central PE, respectively. All analyses were repeated adjusting for age. p-Values of 0.05 and less were considered significant. No adjustment for multiple testing was applied because the statis-
480
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
Table 1 Characteristics of patients separated by peripheral and central pulmonary embolism Peripheral PE
Central PE
Fraction Males Inpatients Dyspnoea Suspected PE Distal-DVTb Femoral-DVTb Iliacal-DVTb
%
45/83 45/83 64/83 69/83 12/45 7/45 11/45
Fraction
54 54 77 84 27 16 24
34/67 34/67 55/67 48/67 13/35 9/35 9/35
Peripheral PE
a b c
53 29 1.1 467 35 12 6
± ± ± ± ± ± ±
%
Un-adjusted
Age-adjusted
51 51 82 73 37 26 26
0.67a
0.54a 0.41a 0.66a 0.27a 0.16a 0.37a 0.68a
0.67a 0.45a 0.09a 0.32a 0.26a 0.90a
Central PE
Mean ± S.D. Age (years) PA-diameter (mm) RV/LV ratio Fibrinogen (mg/dl)b Hematocrit (%)b Hemoglobin (g/dl)b Total protein (g/dl)b
p-Values
16 4 0.3 245 6 2 2
n
Mean ± S.D.
83 83 83 61 65 65 50
62 30 1.2 495 34 11 6
± ± ± ± ± ± ±
p-Values n
13 4 0.4 204 5 2 1
0.0002c 0.12c 0.25c 0.53c 0.52c 0.53c 0.57c
67 67 59 52 60 60 51
– 0.52c 0.53c 0.64c 0.33c 0.35c 0.42c
p-Values derived from logistic regression. Denominators of fractions respective numbers of observations (n) per row do not add to 150 due to missing data (see Section 2). p-Values derived from ANOVA.
tical analysis was performed in an exploratory manner. Odds ratios were estimated from four adjusted multivariate logistic regression models (adjusted for age, age and haemoglobin, age and hematocrit, age and fibrinogen) for central PE and malignancy. Finally an additional adjustment was done including the clinical information (inhospital, dyspnoea, expected PE).
The influence of fibrinogen, hematocrit, hemoglobin and total protein on thrombus density was analysed in a linear regression model. Pearson correlation coefficients were used to exclude a linear dependency. The same procedures were used to prove the effects of these variables on blood density. Pearson correlation coefficients were estimated to exclude correlation of blood and thrombus density. All calculations were
Table 2 Characteristics of patients with cancer and without Variables
Cancer
Non-cancer
Fraction Males Inpatients Dyspnoea Suspected PE Distal-DVTb Femoral-DVTb Iliacal-DVTb Central PEb
%
52/91 53/91 84/91 65/91 12/43 11/43 10/43 47/91
Fraction
57 58 92 72 28 26 23 52
Age (years) PA-diameter (mm) RV/LV ratio Fibrinogen, (mg/dl)b Hematocrit (%)b Hemoglobin (g/dl)b Total protein (g/dl)b Thrombus density (HU)b a b c
59 29 1.1 503 34 11 6 45
± ± ± ± ± ± ± ±
12 4 0.3 251 5 2 2 15
%
27/59 26/59 35/59 52/59 13/37 5/37 10/37 20/59
Cancer Mean ± S.D.
p-Values Un-adjusted
46 44 59 90 35 14 27 34
Mean ± S.D.
91 91 91 68 78 78 67 41
54 29 1.2 446 37 12 6 52
± ± ± ± ± ± ± ±
19 5 0.4 180 6 2 1 13
0.20a 0.13a <0.0001a 0.02a 0.64a 0.24a 0.84a 0.08b
0.09a <0.0001a 0.01a 0.49a 0.18a 0.70a 0.03b
Non-cancer n
Age-adjusted
0.17a
p-Values n 59 59 59 45 47 47 34 16
p-Values derived from logistic regression. Denominators of fractions respective numbers of observations per row do not add to 150 due to missing data (see Section 2). p-Values derived from ANOVA.
0.05c 0.37c 0.94c 0.19c 0.005c 0.006c 0.43c 0.13c
– 0.68c 0.82c 0.24c 0.002c 0.003c 0.50c 0.13c
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
done using the SAS version 9 (SAS Institute, Inc., Cary, NC, USA). 5. Results Table 1 shows the characteristics of patients with central (n = 83) and peripheral PEs (n = 67). Patients with central PE were older than patients with peripheral PEs, but there was no trend to more proximal vein thrombosis in patients with central PEs. The analysed blood parameters did not show any differences. Separating for malignant disease showed that patients with malignant disease were older and had a higher rate of central PEs than patients without malignancies (Table 2). Although dyspnoea was present in most of the cancer patients a lower rate of cancer patients were suspected to have PE compared to patients without cancer. The odds of a central PE in cancer patients was about twice as high as in patients without a malignant disease (Odds ratio: 2.08, 95%-confidence interval: 1.06–4.10; age-adjusted Odds ratio 1.88, 95%-confidence interval: 0.92–3.84; age and haemoglobin adjusted: 2.13 (0.96–4.75); age and hematocrit adjusted: 2.13 (0.95–4.75); age and fibrinogen adjusted: 2.09 (0.93–4.68). After additional adjustment for the clinical information the odds did not deteriorate (Odds ratio: 2.11, 95%-confidence interval: 0.99–4.54; age-adjusted Odds ratio 1.99, 95%-confidence interval: 0.90–4.39; age and haemoglobin adjusted: 2.23 (0.93–5.39); age and hematocrit adjusted: 2.23 (0.93–5.38); age and fibrinogen adjusted: 2.30 (0.95–5.56). In addition cancer patients had significant lower haemoglobin and hematocrit levels. Thrombus density could only be determined in patients with central PE and despite this reduction in total number there was a trend towards a lower thrombus density in patients with malignant disease. Therefore, the influence of the blood parameter on thrombus density was analysed separately. Hematocrit and haemoglobin showed a strong correlation (Pearson correlations coefficient: 0.97447), therefore only haemoglobin was included in a linear regressions model. There is no statistical evidence that the blood parameters haemoglobin, total protein or fibrinogen are related to thrombus density. Thrombus density did not correlate with blood density measured in the perfused pulmonal artery (Pearson correlation coefficient: −0.09590). 6. Discussion The results support the hypothesis that cancer patients are at a higher risk to develop central PEs. Density of the intrapulmonary thrombus seems to differ between patients with cancer and without. The increased risk for PE and PE associated mortality in cancer patients is well known [5,6]. Even the percentage of patients with venous thromboembolic disease at the initial hospitalization is higher for those with cancer compared with those without [7]. Thus in a retrospective examination in a university hospital with high cancer rates the number of patients with cancers with an embolus must be higher. In cancer patients unexpected acci-
481
dental PEs are more frequently found in routine CT scans for staging than in any other group of patients [8–10]. In accordance with these data cancer patients were more frequent in our population, too. The risk of cancer increases with age. Thus the higher age in PE-patients suffering from cancer is not an extraordinary finding and statistical analyses were adjusted for age. There are also some data that reported cancer patients to have a higher rate of massive PE [11]. Fatal PE is one of the major mortality reasons in cancer patients in post-mortem studies [1]. Specific analyses of the rates of central compared to peripheral PEs in cancer patients are not available to our knowledge. Central PE and peripheral PE do not necessarily differ in clot load, because a large embolus floated into the pulmonary artery can be fragmented and cause multiple segmental or subsegmental occlusions. In an older study proximal DVTs have been described to be associated with a higher risk of PE than distal DVT [4]. A more recent study using CT scan on 159 consecutive PE symptom-free patients with symptomatic DVT did not describe a significant association between the level of DVT and the presence of PE [12]. PE recurrences determined in 23 out of 348 patients by lung scan did not depend on proximity of DVT [13]. Even isolated soleal vein thrombosis can be found in patients with PE [14]. All these studies did not distinguish central and peripheral PE. Although we do not have information about the extension of the peripheral DVT our data do not suggest that patients with central PE suffer from more proximal DVT compared to patients with peripheral PE. From a hemodynamical point of view the clinical course of central PE often is more dramatically than in patients with peripheral PE. Although echocardiography is currently used to detect RV dysfunction, CT pulmonary angiography could replace this technique because it can be used to diagnose PE and assess RV function simultaneously. The most sensitive parameter seems to be the ratio of the right and left ventricle (RV/LV ratio). Ghaye et al. reported RV/LV ratios from 1.3 ± 0.4 in PE-survivors to 1.8 ± 0.6 in non-survivors (p = 0.011). Diameter of the pulmonary artery did not distinguish these groups [15]. The RV/LV ratios in this study with patients treated in an intensive care unit were higher than in our population in average and in other studies [16,17]. In these patients with an obviously less hemodynamical deterioration RV/LV ratio is not such sensitive. None of these studies looked for differences in central PE versus peripheral PE which is just an anatomical description. Whereas peripheral PE can have a high clot load, clinically they can be as severe as central PE. Because we did not correlate severity of PE based on clinical findings, we found no difference in RV/LV ration associated to central or peripheral PE. With the direct visualisation of the intrapulmonal embolus multidetector computed tomography allows a precise characterisation of the embolus. HU generally depend on the amount of fluid and cellular structure in the tissue. There is no valid information about HU characteristic for peripheral venous thrombus or even intrapulmonary embolus. The sequential change in density of subdural hematomas (SDHs) in CT has been retrospectively investigated in 446 cases [18]. The study showed that
482
U. Hasenberg et al. / European Journal of Radiology 69 (2009) 478–482
density of acute SDH is usually hyperdense and becomes hypodense within 3 weeks. Then the density progressively increases by the repeated microhemorrhage, which is the mechanism of enlargement of chronic SDH. The density of chronic SDH increases with time up to 90 days, then decreases again after maturation of the neo-membrane, which is the mechanism of spontaneous resolution. In our study the differences in thrombus density could not be explained by cancer specific blood alterations as lower haemoglobin and hematocrit or higher fibrinogen. Thus unknown thrombus age or unknown mechanisms influencing the development of each single thrombus might predict thrombus density. 7. Clinical relevance CT scans can not only detect PE, but can also classify the embolus. The finding of a central PE and a lower thrombus density might represent a specific composition of the thrombus. Such composition might be due to the increased procoagulant activity associated with cancer. Differences in thrombus composition might influence thrombus organisation and solution, respectively, and might influence patient prognosis. Especially the mechanisms causing the differences in size and density of the embolus should be analysed in future studies. 8. Limitations Several limitations in our study warrant consideration. First, its retrospective design prevented us from producing correlations with clinical findings of other diagnostic tests, especially echocardiography. Second, we did not focus on a group of specific symptoms or specific severity. Thus, the rate of central PE can be different in other populations. Third, we did not know the exact age of the DVT. So we cannot correlate thrombus density to thrombus age. Forth, patient selection might have influenced the results. Cancer patients could have had more pronounced clinical symptoms of PE than patients without cancer. Clinical scores such as the Wells score have not been evaluated but would have given crucial information regarding this potential source of error. Although dyspnoea was more frequently in cancer patients the number of patients a PE was suspected in was lower in cancer patients. 9. Conclusion Although this is a retrospective study including a small number of patients it shows different characteristics of PE in cancer and non-cancer patients.
References [1] Lee AY, Levine MN. Venous thromboembolism and cancer: risks and outcomes. Circulation 2003;107:I17–21. [2] Kr¨oger K, Fischer A, Hinrichs A, Gocke C, Rudofsky G. Patients with a malignant tumor have more extended thrombosis than patients without. Angiology 1998;49:923–8. [3] Vucic N, Magdic T, Krnic A, Vcev A, Bozic D. Thrombus size is associated with etiology of deep venous thrombosis—a cross-sectional study. Coll Antropol 2005;29:643–7. [4] Mostbeck A, Partsch H, Kohn H, Konig B. Pulmonary embolism in leg and pelvic vein thrombosis. Results of a prospective study—diagnosis, frequency, nuclearmedical and clinical appearance. Wien Klin Wochenschr 1980;92:464–71. [5] Svendsen E, Karwinski B. Prevalence of pulmonary embolism at necropsy in patients with cancer. J Clin Pathol 1989;42:805–9. [6] Engelke C, Rummeny EJ, Marten K. Pulmonary embolism at multi-detector row CT of chest: one-year survival of treated and untreated patients. Radiology 2006;239:563–75. [7] Levitan N, Dowlati A, Remick SC, et al. Rates of initial and recurrent thromboembolic disease among patients with malignancy versus those without malignancy. Risk analysis using Medicare claims data. Medicine (Baltimore) 1999;78:285–91. [8] Engelke C, Manstein P, Rummeny EJ, Marten K. Suspected and incidental pulmonary embolism on multidetector-row CT: analysis of technical and morphological factors influencing the diagnosis in a cross-sectional cancer centre patient cohort. Clin Radiol 2006;61:71–80. [9] Perrier A, Roy PM, Sanchez O, et al. Multidetector-row computed tomography in suspected pulmonary embolism. N Engl J Med 2005;352:1760–8. [10] Sebastian AJ, Paddon AJ. Clinically unsuspected pulmonary embolism—an important secondary finding in oncology CT. Clin Radiol 2006;61:81–5. [11] Shen VS, Pollak EW. Fatal pulmonary embolism in cancer patients: Is heparin prophylaxis justified? South Med J 1980;73:841–3. [12] Lopez-Beret P, Pinto JM, Romero A, Orgaz A, Fontcuberta J, Oblas M. Systematic study of occult pulmonary thromboembolism in patients with deep venous thrombosis. J Vasc Surg 2001;33:515–21. [13] Monreal M, Lafoz E, Ruiz J, Callejas JM, Arias A. Recurrent pulmonary embolism in patients treated because of acute venous thromboembolism: a prospective study. Eur J Vasc Surg 1994;8(September (5)):584–9. [14] Ohgi S, Tachibana M, Ikebuchi M, Kanaoka Y, Maeda T, Mori T. Pulmonary embolism in patients with isolated soleal vein thrombosis. Angiology 1998;49:759–64. [15] Ghaye B, Ghuysen A, Willems V, et al. Severe pulmonary embolism: pulmonary artery clot load scores and cardiovascular parameters as predictors of mortality. Radiology 2006;239:884–91. [16] Araoz PA, Gotway MB, Trowbridge RL, et al. Helical CT pulmonary angiography predictors of in-hospital morbidity and mortality in patients with acute pulmonary embolism. J Thorac Imaging 2003;18:207–16. [17] van der Meer RW, Pattynama PM, van Strijen MJ, et al. Right ventricular dysfunction and pulmonary obstruction index at helical CT: prediction of clinical outcome during 3-month follow-up in patients with acute pulmonary embolism. Radiology 2005;235:798–803. [18] Lee KS, Bae WK, Bae HG, Doh JW, Yun IG. The computed tomographic attenuation and the age of subdural hematomas. J Korean Med Sci 1997;12:353–9.